A network-centric input-output and robustness analysis framework for distributed dynamic systems

University Of Washington, Seattle WA

Investigators

Abstract

Research Plan: A networked dynamic system is a collection of dynamical units that interact via an information exchange network during their operation. These systems have found many applications in diverse areas such as multiple space, air, and land vehicles, energy systems, physiology, and medicine. Currently, there is an active research effort underway in the control and systems community to formalize these systems and lay out a foundation for their analysis and synthesis. The objective of this project is to carry out foundational studies on a network-centric input-output and robustness analysis framework for distributed dynamic systems. The approach adopted for this purpose is based on algebraic graph theory on one hand, and random and random geometric graphs, on the other. In this direction, we will address problems associated with probabilistic moments of system-theoretic objective functionals over uncertain networks, as well as the important phenomena associated with time-scales and phase transitions in robustness and performance measures of the corresponding networked systems. An existing computational testbed at the University of Washington will be augmented for the purpose of validation and evaluation of the proposed methodologies. Education Plan: Distributed dynamic systems provide an ideal venue to engage control and systems graduate students in the elegance and power of system ideas, not only as they pertain to engineering problems, but also in their relation to biology and information sciences. The education program of this project is implemented through a graduate course offering and curriculum development, graduate mentoring, and promoting research collaboration venues via in-class seminars and a workshop. The proposed course, Networkcentric Dynamic System Theory and Applications, will be developed and offered as a two quarters graduate level course in the controls curriculum of the University of Washington. The course will cover aspects of combinatorial theory that, in conjunction with system theory, have proved instrumental in the analysis of network-enabled dynamic systems. The second quarter of the course provides a rigorous bridge between system theory and networked dynamic systems as encountered in a wide array of disciplines such as in biology and physiology. The workshop on network-centric dynamic systems during the final year of the project will not only serve as an exceptional educational outlet for the graduate students, but also as a vehicle to further enhance opportunities for interdisciplinary research. Intellectual Merit: The intellectual merit of the proposed research and education program consists of: (1) foundational contributions to a new sub-discipline in control theory referred to as network-centric dynamic system theory (2) applying novel graph and system theoretic techniques for the analysis of networked dynamic systems as encountered in life and physical sciences and engineering systems (3) providing a significant impetus for collaborations between system and control theorists, combinatorialists, biologists, and physicists, at the University of Washington, and (4) development of a new curriculum for graduate program in systems and control education. Broader Impact: Our project will lay a fertile ground work for future collaborations between system and control researchers, graph theorists, and the biological science community. By doing so, it will also bring into the horizon a host of new career options for system and control graduate students. Our contributions to the emerging field of networks will be well publicized via: (1) a well-organized workshop on the third year of the project, bringing researchers from across disciplines, from local computer science, engineering, and biomedical industries (2) presenting research results, often by graduate students, in professional conferences and research institutions, and (3) frequent seminars in the educational institutions that are non-Ph.D granting and/or serve underrepresented groups. These more public seminars will highlight the common complexities of networked dynamic systems across various disciplines, and how control and system theory is providing novel theoretical and algorithmic tools for their analysis and synthesis. These seminars will also be actively used as a recruitment tool, particularly for attracting potential graduate students from underrepresented groups in engineering.